envopts.c revision 147493
1/*- 2 * Copyright (c) 2005 - Garance Alistair Drosehn <gad@FreeBSD.org>. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 * 26 * The views and conclusions contained in the software and documentation 27 * are those of the authors and should not be interpreted as representing 28 * official policies, either expressed or implied, of the FreeBSD Project. 29 */ 30 31#include <sys/cdefs.h> 32__FBSDID("$FreeBSD: head/usr.bin/env/envopts.c 147493 2005-06-20 03:43:25Z gad $"); 33 34#include <sys/stat.h> 35#include <sys/param.h> 36#include <err.h> 37#include <errno.h> 38#include <ctype.h> 39#include <stdio.h> 40#include <stdlib.h> 41#include <string.h> 42#include <unistd.h> 43 44#include "envopts.h" 45 46static void expand_vars(char **thisarg_p, char **dest_p, const char 47 **src_p); 48static int is_there(char *candidate); 49 50/* 51 * The is*() routines take a parameter of 'int', but expect values in the range 52 * of unsigned char. Define some wrappers which take a value of type 'char', 53 * whether signed or unsigned, and ensure the value ends up in the right range. 54 */ 55#define isalnumch(Anychar) isalnum((u_char)(Anychar)) 56#define isalphach(Anychar) isalpha((u_char)(Anychar)) 57#define isspacech(Anychar) isspace((u_char)(Anychar)) 58 59/* 60 * Routine to determine if a given fully-qualified filename is executable. 61 * This is copied almost verbatim from FreeBSD's usr.bin/which/which.c. 62 */ 63static int 64is_there(char *candidate) 65{ 66 struct stat fin; 67 68 /* XXX work around access(2) false positives for superuser */ 69 if (access(candidate, X_OK) == 0 && 70 stat(candidate, &fin) == 0 && 71 S_ISREG(fin.st_mode) && 72 (getuid() != 0 || 73 (fin.st_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) != 0)) { 74 if (env_verbosity > 1) 75 fprintf(stderr, "#env matched:\t'%s'\n", candidate); 76 return (1); 77 } 78 return (0); 79} 80 81/** 82 * Routine to search through an alternate path-list, looking for a given 83 * filename to execute. If the file is found, replace the original 84 * unqualified name with a fully-qualified path. This allows `env' to 85 * execute programs from a specific strict list of possible paths, without 86 * changing the value of PATH seen by the program which will be executed. 87 * E.G.: 88 * #!/usr/bin/env -S-P/usr/local/bin:/usr/bin perl 89 * will execute /usr/local/bin/perl or /usr/bin/perl (whichever is found 90 * first), no matter what the current value of PATH is, and without 91 * changing the value of PATH that the script will see when it runs. 92 * 93 * This is similar to the print_matches() routine in usr.bin/which/which.c. 94 */ 95void 96search_paths(char *path, char **argv) 97{ 98 char candidate[PATH_MAX]; 99 const char *d; 100 char *filename, *fqname; 101 102 /* If the file has a `/' in it, then no search is done */ 103 filename = *argv; 104 if (strchr(filename, '/') != NULL) 105 return; 106 107 if (env_verbosity > 1) { 108 fprintf(stderr, "#env Searching:\t'%s'\n", path); 109 fprintf(stderr, "#env for file:\t'%s'\n", filename); 110 } 111 112 fqname = NULL; 113 while ((d = strsep(&path, ":")) != NULL) { 114 if (*d == '\0') 115 d = "."; 116 if (snprintf(candidate, sizeof(candidate), "%s/%s", d, 117 filename) >= (int)sizeof(candidate)) 118 continue; 119 if (is_there(candidate)) { 120 fqname = candidate; 121 break; 122 } 123 } 124 125 if (fqname == NULL) { 126 errno = ENOENT; 127 err(127, "%s", filename); 128 } 129 *argv = strdup(candidate); 130} 131 132/** 133 * Routine to split a string into multiple parameters, while recognizing a 134 * few special characters. It recognizes both single and double-quoted 135 * strings. This processing is designed entirely for the benefit of the 136 * parsing of "#!"-lines (aka "shebang" lines == the first line of an 137 * executable script). Different operating systems parse that line in very 138 * different ways, and this split-on-spaces processing is meant to provide 139 * ways to specify arbitrary arguments on that line, no matter how the OS 140 * parses it. 141 * 142 * Within a single-quoted string, the two characters "\'" are treated as 143 * a literal "'" character to add to the string, and "\\" are treated as 144 * a literal "\" character to add. Other than that, all characters are 145 * copied until the processing gets to a terminating "'". 146 * 147 * Within a double-quoted string, many more "\"-style escape sequences 148 * are recognized, mostly copied from what is recognized in the `printf' 149 * command. Some OS's will not allow a literal blank character to be 150 * included in the one argument that they recognize on a shebang-line, 151 * so a few additional escape-sequences are defined to provide ways to 152 * specify blanks. 153 * 154 * Within a double-quoted string "\_" is turned into a literal blank. 155 * (Inside of a single-quoted string, the two characters are just copied) 156 * Outside of a quoted string, "\_" is treated as both a blank, and the 157 * end of the current argument. So with a shelbang-line of: 158 * #!/usr/bin/env -SA=avalue\_perl 159 * the -S value would be broken up into arguments "A=avalue" and "perl". 160 */ 161void 162split_spaces(const char *str, int *origind, int *origc, char ***origv) 163{ 164 const char *bq_src, *src; 165 char *dest, **newargv, *newstr, **nextarg, **oldarg; 166 int addcount, bq_destlen, copychar, found_sep, in_arg, in_dq, in_sq; 167 168 /* 169 * Ignore leading space on the string, and then malloc enough room 170 * to build a copy of it. The copy might end up shorter than the 171 * original, due to quoted strings and '\'-processing. 172 */ 173 while (isspacech(*str)) 174 str++; 175 if (*str == '\0') 176 return; 177 newstr = malloc(strlen(str) + 1); 178 179 /* 180 * Allocate plenty of space for the new array of arg-pointers, 181 * and start that array off with the first element of the old 182 * array. 183 */ 184 newargv = malloc((*origc + (strlen(str) / 2) + 2) * sizeof(char *)); 185 nextarg = newargv; 186 *nextarg++ = **origv; 187 188 /* Come up with the new args by splitting up the given string. */ 189 addcount = 0; 190 bq_destlen = in_arg = in_dq = in_sq = 0; 191 bq_src = NULL; 192 for (src = str, dest = newstr; *src != '\0'; src++) { 193 copychar = found_sep = 0; 194 switch (*src) { 195 case '"': 196 if (in_sq) 197 copychar = *src; 198 else if (in_dq) 199 in_dq = 0; 200 else { 201 in_dq = 1; 202 bq_destlen = dest - *(nextarg - 1); 203 bq_src = src; 204 } 205 break; 206 case '$': 207 if (in_sq) 208 copychar = *src; 209 else { 210 expand_vars((nextarg - 1), &dest, &src); 211 } 212 break; 213 case '\'': 214 if (in_dq) 215 copychar = *src; 216 else if (in_sq) 217 in_sq = 0; 218 else { 219 in_sq = 1; 220 bq_destlen = dest - *(nextarg - 1); 221 bq_src = src; 222 } 223 break; 224 case '\\': 225 if (in_sq) { 226 /* 227 * Inside single-quoted strings, only the 228 * "\'" and "\\" are recognized as special 229 * strings. 230 */ 231 copychar = *(src + 1); 232 if (copychar == '\'' || copychar == '\\') 233 src++; 234 else 235 copychar = *src; 236 break; 237 } 238 src++; 239 switch (*src) { 240 case '"': 241 case '#': 242 case '$': 243 case '\'': 244 case '\\': 245 copychar = *src; 246 break; 247 case '_': 248 /* 249 * Alternate way to get a blank, which allows 250 * that blank be used to separate arguments 251 * when it is not inside a quoted string. 252 */ 253 if (in_dq) 254 copychar = ' '; 255 else { 256 found_sep = 1; 257 src++; 258 } 259 break; 260 case 'c': 261 /* 262 * Ignore remaining characters in the -S string. 263 * This would not make sense if found in the 264 * middle of a quoted string. 265 */ 266 if (in_dq) 267 errx(1, "Sequence '\\%c' is not allowed" 268 " in quoted strings", *src); 269 goto str_done; 270 case 'f': 271 copychar = '\f'; 272 break; 273 case 'n': 274 copychar = '\n'; 275 break; 276 case 'r': 277 copychar = '\r'; 278 break; 279 case 't': 280 copychar = '\t'; 281 break; 282 case 'v': 283 copychar = '\v'; 284 break; 285 default: 286 if (isspacech(*src)) 287 copychar = *src; 288 else 289 errx(1, "Invalid sequence '\\%c' in -S", 290 *src); 291 } 292 break; 293 default: 294 if ((in_dq || in_sq) && in_arg) 295 copychar = *src; 296 else if (in_arg && isspacech(*src)) 297 found_sep = 1; 298 else { 299 /* 300 * If the first character of a new argument 301 * is `#', then ignore the remaining chars. 302 */ 303 if (!in_arg && *src == '#') 304 goto str_done; 305 copychar = *src; 306 } 307 } 308 if (copychar) { 309 if (!in_arg) { 310 /* This is the first byte of a new argument */ 311 *nextarg++ = dest; 312 addcount++; 313 in_arg = 1; 314 } 315 *dest++ = (char)copychar; 316 } else if (found_sep) { 317 *dest++ = '\0'; 318 while (isspacech(*src)) 319 src++; 320 --src; 321 in_arg = 0; 322 } 323 } 324str_done: 325 *dest = '\0'; 326 *nextarg = NULL; 327 if (in_dq || in_sq) { 328 errx(1, "No terminating quote for string: %.*s%s", 329 bq_destlen, *(nextarg - 1), bq_src); 330 } 331 if (env_verbosity > 1) { 332 fprintf(stderr, "#env split -S:\t'%s'\n", str); 333 oldarg = newargv + 1; 334 fprintf(stderr, "#env into:\t'%s'\n", *oldarg); 335 for (oldarg++; *oldarg; oldarg++) 336 fprintf(stderr, "#env &\t'%s'\n", *oldarg); 337 } 338 339 /* Copy the unprocessed arg-pointers from the original array */ 340 for (oldarg = *origv + *origind; *oldarg; oldarg++) 341 *nextarg++ = *oldarg; 342 *nextarg = NULL; 343 344 /* Update optind/argc/argv in the calling routine */ 345 *origind = 1; 346 *origc += addcount; 347 *origv = newargv; 348} 349 350/** 351 * Routine to split expand any environment variables referenced in the string 352 * that -S is processing. For now it only supports the form ${VARNAME}. It 353 * explicitly does not support $VARNAME, and obviously can not handle special 354 * shell-variables such as $?, $*, $1, etc. It is called with *src_p pointing 355 * at the initial '$', and if successful it will update *src_p, *dest_p, and 356 * possibly *thisarg_p in the calling routine. 357 */ 358void 359expand_vars(char **thisarg_p, char **dest_p, const char **src_p) 360{ 361 const char *vbegin, *vend, *vvalue; 362 char *edest, *newstr, *vname; 363 int bad_reference; 364 size_t namelen, newlen; 365 366 bad_reference = 1; 367 vbegin = vend = (*src_p) + 1; 368 if (*vbegin++ == '{') 369 if (*vbegin == '_' || isalphach(*vbegin)) { 370 vend = vbegin + 1; 371 while (*vend == '_' || isalnumch(*vend)) 372 vend++; 373 if (*vend == '}') 374 bad_reference = 0; 375 } 376 if (bad_reference) 377 errx(1, "Only ${VARNAME} expansion is supported, error at: %s", 378 *src_p); 379 380 /* 381 * We now know we have a valid environment variable name, so update 382 * the caller's source-pointer to the last character in that reference, 383 * and then pick up the matching value. If the variable is not found, 384 * or if it has a null value, then our work here is done. 385 */ 386 *src_p = vend; 387 namelen = vend - vbegin + 1; 388 vname = malloc(namelen); 389 strlcpy(vname, vbegin, namelen); 390 vvalue = getenv(vname); 391 if (vvalue == NULL || *vvalue == '\0') { 392 if (env_verbosity > 2) 393 fprintf(stderr, 394 "#env replacing ${%s} with null string\n", 395 vname); 396 return; 397 } 398 399 if (env_verbosity > 2) 400 fprintf(stderr, "#env expanding ${%s} into '%s'\n", vname, 401 vvalue); 402 403 /* 404 * There is some value to copy to the destination. If the value is 405 * shorter than the ${VARNAME} reference that it replaces, then we 406 * can just copy the value to the existing destination. 407 */ 408 edest = *dest_p; 409 if (strlen(vname) + 3 >= strlen(vvalue)) { 410 while (*vvalue != '\0') 411 *edest++ = *vvalue++; 412 *dest_p = edest; 413 return; 414 } 415 416 /* 417 * The value is longer than the string it replaces, which means the 418 * present destination area is too small to hold it. Create a new 419 * destination area, copy the present 'thisarg' value and the value 420 * of the referenced-variable to it, and then update the caller's 421 * 'thisarg' and 'dest' variables to match. 422 */ 423 *edest = '\0'; /* Provide terminator for 'thisarg' */ 424 newlen = strlen(*thisarg_p) + strlen(vvalue) + strlen(*src_p) + 1; 425 newstr = malloc(newlen); 426 strcpy(newstr, *thisarg_p); 427 strcat(newstr, vvalue); 428 *thisarg_p = newstr; 429 *dest_p = strchr(newstr, '\0'); 430} 431